Phase 2 - Coordination Dividend Simulation

Vertical-NORC prototype on synthetic operational data. Each building's visits are packed into the minimum number of worker-shifts with an OR-Tools CP-SAT bin-packing model. Siloed packs every agency separately (today); coordinated pools the building's visits. Inputs: PSW org counts and service hours per building. All else is a tunable planning assumption (shift 6.0 h, overhead 0.75 h/dispatch, 312 operating days/yr).
448Multi-agency buildings modeled
1682 → 1172Worker-shifts/day (siloed → coordinated)510 fewer dispatches/day
30%Fewer workers entering buildingsutil 57% → 82%
119,340 hCoordination dividend / year~74.6 FTE of pure overhead
4.5 → 3.0Distinct workers/week per building (continuity)33% fewer faces per client

Where coordination frees up the most workers

Siloed (per-agency) Coordinated (pooled)
980 Thornbury Lane S7→2Building 0016→2Building 00210→6Building 00310→6Building 0045→110 Old York Mills Rd5→1Building 00511→7Building 0065→2243 Bramwell Lane N9→6Building 0076→3Building 0088→5Building 0097→4
Each pair of bars is one building: red = workers dispatched today (each agency sends its own), teal = workers needed if the building's visits were pooled. The gap is eliminated travel-to-site and idle time - the dividend that routing optimization cannot reach in a vertical NORC.

Inside the packing - showcase buildings

980 Thornbury Lane S Condos
8 agencies · 12 visits · 12.0 visit-h/day · siloed 7 workers → coordinated 2 workers
Coordinated packing (each bar = one worker's day):
W1W2
Building 001 Community Housing
6 agencies · 12 visits · 11.0 visit-h/day · siloed 6 workers → coordinated 2 workers
Coordinated packing (each bar = one worker's day):
W1W2
Building 002 Condos
7 agencies · 31 visits · 32.5 visit-h/day · siloed 10 workers → coordinated 6 workers
Coordinated packing (each bar = one worker's day):
W1W2W3W4W5W6
Each bar is one coordinated worker's day; coloured segments are visits stacked against the 6.0-hour shift capacity (grey). High fill = high utilization. Under the siloed model these visits would be scattered across 8+ agencies' separate workers.

Top buildings by dividend

BuildingTypeAgenciesVisits/day Visit-h/daySiloedCoord.Workers saved Utilization
980 Thornbury Lane SCondos81212.072529% → 100%
Building 001Community Housing61211.062431% → 92%
Building 002Condos73132.5106454% → 90%
Building 003Condos83434.0106457% → 94%
Building 004Condos555.051417% → 83%
10 Old York Mills RdCondos676.051420% → 100%
Building 005Apartments73940.5117461% → 96%
Building 006Community Housing5911.552338% → 96%
243 Bramwell Lane NCommunity Housing43035.096365% → 97%
Building 007Community Housing61214.063339% → 78%
Building 008Condos62528.585359% → 95%
Building 009Condos42123.574356% → 98%
Building 010Condos51112.052340% → 100%
5 Northwood Lane NCommunity Housing72024.585351% → 82%
215 The Donway WCondos465.041321% → 83%
205 The Donway WCondos61213.563338% → 75%
Building 011Community Housing61011.552338% → 96%
Building 012Condos61717.063347% → 94%
Building 013Apartments61515.563343% → 86%
16 Quarry Lane NCommunity Housing52118.574344% → 77%
Building 014Community Housing51114.063339% → 78%
Building 015Community Housing51620.074348% → 83%
Building 016Apartments589.052330% → 75%
115 The EsplanadeCo-Ops52021.574351% → 90%
Building 017Apartments61817.063347% → 94%
Caveat: with no client time-windows or durations, this models structural consolidation (how few workers could cover the demand), not minute-by-minute schedule feasibility. It upgrades to full VRPTW once real availability/time-window data arrives - the same CP-SAT engine, with travel and time-window constraints added.